GPS Guided Autonomous Rover

ELEN3017

Josh Berman 320856

Group: 16

Partners: Terry Bugai, Kent Salmon, Mayur Chiba

Date: 24 May 2013

School of Electrical & Information Engineering,

University of the Witwatersrand,
Private Bag 3,
2050,
Johannesburg,
South Africa
Abstract
This document provides the development and implementation of a GPS (Global Positioning System) guided rover.
The rover is to compete in a race on a pre-defined track consisting of vertices made up of GPS coordinates. The
design is comprised of a perspex base, battery pack, servo motors, Ublox GPS module and an ATMega328 Arduino
microcontroller. The microcontroller receives the input from the GPS module in string format through serial
communication. This data is then sent through our path finding algorithm and the direction is calculated. The rover
runs through the course with minor mishaps. Improvement recommendations include alternate algorithms and
additional sensors.

1. INTRODUCTION The battery pack must be the sole power supply

Critical to this project, a Ublox GPS unit is mounted on
the highest point on the rovers surface. Location data is
converted to a usable form and a path is then calculated
using our path finding Algorithm. The rover determines
its location and direction in order to estimate how many
degrees to rotate. It then rotates accordingly, drives
forward and re-calibrates. This process continues until the
rover has reached its destination.
Section 2 defines the constraints versus requirements of
this autonomous rover. The physical circuit
implementation is discussed in section 3 and the
construction and implementation in section 4. The
stability of the physical structure of the rover is discussed
in section 5, the algorithm is discussed in section 6 with
all results in section 7. The concluding statements with
recommendations appear in section 8.
for the rover. The rover must hence be able to
provide enough torque to carry the batter pack.
The Ublox GPS module must be used to control
the direction of the rover.
The Perspex provided must be used in the
construction of the rover
No power tools are to be used throughout the
construction of the rover.
2.2. Success Criteria
Navigate through a track consisting of vertices
made of GPS coordinates in the quickest
possible time without veering off course.
2.3. Component Constraints

For the purposes of this autonomous rover, a standardised

2. CONSTRAINTS AND REQUIREMENTS
2.1. General Assumption and Constraints
The continuous servo motors given must be used
to drive and steer the rover.
issuing of components to be used as power supplies and
tracking modules forms the basic constraints of the
subsequent system. It is around these components in table
1, that the subsequent design must meet all the
requirements.

1
Table 1: Issued components and relevant ratings Table 2: Tested load characteristics of rover under normal
conditions.
Components Relevant
Ratings Component Voltage(V) Load Load
Current(A) Power
2 Continuous Rotation Servo 6V (W)
Motors Arduino 5.0 80m 0.4
1 Battery pack 6V 2.8Ah GPS 3.3 40m 0.13
50 300mm Perspex sheet N/A Servos 6.0 190m 1.14
2 Rubber/plastic wheels (not 9cm diameter Rover in motion 6.0 360m 2.16
issued) (under no load)
1 Track ball (not issued) N/A Rover in motion 6.0 420m 2.52
(under load)

3. DESIGN APPROACH
3.1. General Design Approach
When comparing the constraints and requirements of the
design the following two options for the drive mechanism
of the rover are presented. One way is to use a single
continuous rotation servo to produce the driving force
acting on the rear axle of the rover and to use a 0-180
semi rotational servo to direct the front wheel which
controls the direction. Another way is to use two
continuous rotation servos to drive each wheel
independently. This would allow us to turn on our axis by
rotating the wheels in opposite directions. The latter
option became the chosen one as the rover needs to make
immediate turns when advancing passed coordinates.
The choice of the microprocessor used is dependent on a
number of factors. The microprocessor must contain
enough memory to house our entire code. The processing
speed must be enough to calculate in real time before the
rover crosses the next point. The micro must contain
enough input/output pins so that the GPS and servo
motors can be connected.
The Arduino ATMega328 fits all this criteria and is
readily available at most electronic suppliers. The
sketch is coded in c providing an easier platform than
PICs requiring assembly or mipps.
Load currents are tested using a DC power supply and
reading off the current drawn. These test results show that
the rover requires at least a 6V supply at <1Amp.
3.3. Servo Motors
Two FS5109R standard continuous rotation servos are
used to drive and steer the rover. A PWM signal
generated by the microprocessor controls the angle of
rotation. Since our servos are continuous rotation, the
angle is used to determine the wheel direction once off,
i.e. forwards or backwards. The servos continue to rotate
until a new command is given as they do not contain a
limiting potentiometer. In order to turn the rover, the
servos must be set to rotate the wheels in opposite
directions. The duty cycle of the PWM determines the
direction of the rotation of the servos and the frequency
determines the speed of rotation. Our frequency is set in
order to get maximum speed of rotation while still
providing enough torque to carry the battery pay load.
This frequency is set to be 333.3 Hz rated at the
maximum our plastic gears can handle. The normal
operational frequency is between 50-60 Hz [4]. The
rovers speed is noticeably faster.
Table 3: Servo Motor Ratings from datasheet
Components Ratings Ratings

ASSEMBLY AND PHYSICAL FEATURES Power 4.8V 6V

Speed 0.16sec/60 0.16sec/60
3.2. Power supply Torque 6.5kg/cm 9.5kg/cm
Weight 52g N/A
The Battery pack provided has an output voltage Size 40.8*20.1*38.0mm N/A
measured to be 6.3V when fully charged. This is the sole Application robot N/A
supply for the rover and must provide enough current to
power the servo motors, GPS unit and Arduino
microprocessor. The optimum voltage for maximum
torque of the servos is 7.2V [1], but as little as a 4V 3.4. The Wheels
supply can be used. The GPS module requires 3.3V [2]
and is regulated through the Arduino. The microprocessor The diameter of the wheels is a crucial factor to the speed
requires a 5V supply but anything up to 8V can be used as of the rover. As the servos are now rotating at a fixed
the PCB board contains its own 5Vdc regulator [3]. These speed, Assuming enough torque is provided through
values are measured experimentally using a controlled entire operation, we can assume the larger the wheels the
variable power supply and displayed in table 2 as greater the speed. This is limited to availability of
minimum values for acceptable levels of performance. attachable wheels and as mentioned the servo torque.

2
Different sets of wheels are tried and tested until the load
on the servos is minimised and the speed of the rover is
maximised. A final wheel of diameter 9cm is chosen. The
wheels are taken from an old Mecano/Lego set and are
fixed in place using rods. Glue is used to secure these
rods. A table of varying size wheels and their respective
speed is shown below.
Table 4: Wheel diameter and resultant speed/time
diameter Perimeter Rotations Time to
per meter cover a meter
(sec/meter)
4cm 12.5cm 8 7.68
6cm 18.85cm 5.3 5.08
9cm 28.27cm 3.53 3.39
Speed is the Inverse of the time taken to cover a meter.
3.5. The Chassis
A 6mm thick sheet of Perspex measure to be 300x500mm
is issued and must contribute to the structure of the rover.
The Perspex is light in comparison to a metal structure
but heavy when considering the added weight of the
battery pack. Perspex is used as it can be easily cut using
a saw or bored using a hand drill. Cut out pieces can be
attached using a glue gun. Aesthetic additions such as a
wind deflector are added for our own amusement. The
Chassis is a single sheet of Perspex with side sections
removed for the wheels to fit. Holes are drilled for the
Arduino to be secured and the front track ball to be
mounted. The base plate dimensions can be found in
Appendix A, Figure 1.
3.6. GPS Module
The Global Positioning System comprises of multiple
Satellite networks orbiting the earth. These satellites
transmit data or timestamps depending on their position
in space. The data is received by a GPS receiver where
there is an unobstructed line of sight. This is accessible to
anyone from military personnel to civilians. The most
common GPS receivers are located in land mapping
devices used in cars and cell phones [5].
The data set or information from GPS receiver is called
NMEA, which stands for National Marine Electronics
Association. These data strings based on ASCII are
communicated at a rate of 38400 bits per second [6]
which is equivalent to the baud rate of 38400 characters
per-second. The data is transmitted in sentence codes and
is decoded into string format which we then manipulate to
find our required information. The GPS data is normally
received and transmitted in a standard NMEA-0183
format. This GPS continuously outputs NMEA sentence
codes such as RMC and VTG. In this case, the only
concerned is the ($GPRMC) sentence, which represents
the Global Positioning Recommended Minimum Specific
GPS/Transit Data.
A serial communication port is set up in the software of
3
the Arduino in order to receive the desired data. Once this
data is received we can use string manipulation to extract
the latitude and longitude. Checks need to be performed
in order to make sure data is being received else the
information is insufficient.
The GPS module takes roughly 30 seconds to begin
receiving information and therefore initial loops need to
be put in place to prevent the rover from moving forward
before this information is received. The Latitude and
longitude are displayed with five decimal places in order
to depict an accuracy of 1.1 meters [6]. The accuracy
during final tests was less than this and caused minor path
variations for this rover. With a larger antenna and the
storage of latitude and longitude in integer format as
opposed to floats, we can reduce this error significantly.
There was not enough time to incorporate this fix in the
code and hence the error was relatively high. The GPS
output can be found in Appendix B.
4. CONSTRUCTION
The rover is pieced together using assorted screws and a
glue gun. The Servos are attached in opposite directions
to the underside of the Perspex base plate. The wheels are
then attached independently to each servo using the
screws and the servo arms that are supplied with the
servos. The track ball is fixed using the holes in the front
of the chassis that were drilled previously. The battery
pack is strategically placed and secured in the center of
the chassis for optimum weight distribution. The GPS unit
is placed on top of the battery pack and the TX and RX
pins are connected to pin 3 and 5 of the Arduino
respectively. The Vcc and GND are connected directly to
the 3.3V output of the Arduino and the GND pin of the
Arduino respectively. The Arduino is secured in front of
the battery pack and the supply voltage is connected
directly, via a power diode, into the power connector. The
servo signal inputs are connected to PWM pins 6 and 10
of the Arduino and their respective ground and supply are
connected to the battery supply rail. The rear spoiler is
added for aesthetic reasons, and a rear leg is then secured.
The construction of the rover is complete and the Arduino
sketch containing the path finding algorithm can be
uploaded.
5. DESIGN INSTABILITY
The performance of the rover is heavily susceptible to
wind resistance and uneven surfaces. The small trackball
in the front can get lodged in a crack in the surface of the
track thus altering the direction of movement. More
calculations need to be done in order to get the rover back
on track as it is no longer moving along its desired path.
The Battery pack is large and heavy and puts a great deal
of strain on the servos. The battery is located slightly
forward on the chassis. Ideally one would like to have a
50:50 weight distribution but as a result of the vibration
caused by the uneven surface, the weight is constantly
shifted and the rover can topple backwards. A support bar
 protruding from the rear of the rover prevents the rover 7. RESULTS
from toppling backwards if the weight is shifted
drastically. 7.1. Test day

The wheels need to be perfectly aligned as a slight degree
of inaccuracy can cause a large change in direction when
travelling over large distances. This takes copious trials
and error. Additionally the servos can be tweaked by
altering their variable resistors; this can slow down a
servo independently allowing the rover to tend towards a
certain direction. This is used for correction of alignment
and compensation of independent speed inaccuracy.
6. PATH FINDING ALGORITHM
The path finding algorithm is initially drafted in sudo
code to get a better understanding of the objective
required. The code is then simulated in Matlab which
provides a visual representation of the calculated angles
and their respective paths. These simulated results can be
seen in Appendix C. Once the algorithm is tried and
tested on Matlab, it is then coded in c and used on the
Arduino to output the required results.
The algorithm begins with finding the direction required
for the rover to move. This can only be done with two sets
of reference coordinates. Thus the rover must move
forward initially to retrieve these two sets of coordinates.
The coordinates are passed to a function that calculates
the angle of movement and the rover is rotated
accordingly. The rover must continuously loop through
this set of instructions until the desired point is reached.
The rover then checks for the next vertices and re-
evaluates its position.
Assumptions are made for the testing of the rover: the
course is assumed to be smooth, the GPS module would
constantly provide accurate information and that the
power supply would be constant. This is an ideal case.
During testing however these factors come into play and
cause variations in the results. The rover completed the
course but overshot one or two vertices on occasion. The
speed and direction was hindered by the uneven surface
which caused the rover to veer off the path towards the
end. After much initial tests prior to the demonstration,
the battery output voltage was lower than its initial value
and the rover had to be connected directly to a PC in
order to provide enough current for all the components.
The rover did manage to complete the entire path at a
relatively fast pace with all the constraints adhered to.
8. IMPROVEMENTS
The rovers improvements can be made in the form of
additional sensors such as sonar or a gyroscope. The
sonar sensor can measure the distance to an object in its
path and hence trigger an object avoidance algorithm. An
example of this algorithm is shown in Appendix D. The
gyroscope can be used to measure the angle of rotation.
This would be preferable as the current design requires
the rover to move forward and retrieve new coordinates
and hence calculate its angle.
The rover build quality was of high standard but the
uneven surface can cause undesired results. The rover can
be built with a wider base and thicker tires to absorb
bumps in the track.

An antenna can be added to improve the GPS signal and

therefore reduce the time taken to retrieve coordinates;
this would noticeably speed up the rovers algorithm. To
add theses improvements the connection diagram in
Appendix E can be used for reference.

9. CONCLUSION

A GPS guided autonomous rover was designed, constructed and tested. The system followed a strict set of constraints
with regards to the structure of the Rover as well as the implementation. The employed system autonomously navigates
its way through a path set out in GPS coordinates powered by a single 6.3V battery pack. The accuracy of the GPS
module was questioned and the desired outcome was hindered. The algorithm that was initially drafted in Matlab did
not function as well as expected when converted to c, but the overall construction and implementation produced the
required result. This project is a prime example of an intelligent autonomous vehicle.

4
 REFERENCES

[1] Servo Voltage, FS5109R Data-sheet.

[2] GPS Operating Voltage, http://www.openimpulse.com/blog/products-page/product-category/ublox-neo-6m-gps-
module/ ,June 2013
[3] Arduino Operating Voltage , http://arduino.cc/en/Main/arduinoBoardUno, June 2013
[4] Servo Frequency, http://pcbheaven.com/wikipages/How_RC_Servos_Works/ , June 2013.
[5] GPS information, http://en.wikipedia.org/wiki/Global_Positioning_System , May 2013.
[6] Implementation of GPS for Location Tracking , 2011 IEEE Control and System Graduate Research Colloquium,
Shah Alam, Malaysia

Appendix A

Figure 1: Perspex Chassis Dimensions

Figures 2: Side View of Rover Figures 3: Top View of Rover

5
 Figures 4: Rear View of Rover

Appendix B

Figure 1: GPS Output Stream Using Serial Communication

Appendix C

Figure 1: Flow Diagram of the Path Finding Algorithm

6
 Sudo code:

General order
Initialisation
Find goal
drive forward
determine "direction"
goalFind = 0
Running
turn & drive
if goalFind = 1
find goal + 1
goalFind =0
end
acquire position
check angle
check if success
if success
goal = goal + 1
goalFind = 1
end

Issue that may arise in order of probability:

###
Check angle after movement, else every turn/drive adds an error
###
Crossing goal latitude/longitude
if condition: latP > latS
if latP >= latG
latG = max(latC,latB)
end
if condition: latP < latS
if latP <= latG
longG = min(latC,latB)
end
similarly for longitude.
###
Crossing wall
Check proximity to wall AC

Figure 2: Map coordinates illistration

7
 Figure 3: Visual Matlab Map Simulation

MATLAB code:

clc;
%close all;
%clear all;

latIn = [28.02766, 28.02783, 28.02778, 28.02778, 28.02785, 28.02763];

longIn = [26.19165, 26.19164, 26.19159, 26.19149, 26.19139, 26.19150];
latOut = [28.02767, 28.02794, 28.02792, 28.02787, 28.02792 ,28.02798, 28.02765];
longOut = [26.19173, 26.19166, 26.19155, 26.19155, 26.19155, 26.19131,
26.19141];
latP= 28.02767;
longP = 26.19169;
longVert = [26.19165, 26.19173, 26.19164, 26.19166, 26.19155, 26.19155,
26.19159, 26.19149, 26.19155, 26.19139, 26.19131, 26.19141, 26.19150];
latVert = [28.02766, 28.02767, 28.02783, 28.02794, 28.02792, 28.02787, 28.02778,
28.02778, 28.02792, 28.02785, 28.02798, 28.02765, 28.02763];

longPath = longP;
latPath = latP;
longA = longVert(1,1);
latA = latVert(1,1);
longB = longVert(1,2);
latB = latVert(1,2);
longC = longVert(1,3);
8
latC = latVert(1,3);
longD = longVert(1,4);
latD = latVert(1,4);

longABC = [longA, longB, longC , longA];

latABC = [latA , latB , latC , latA ];
longPC = [longP, longC];
latPC = [latP , latC ];

if (longA - longD)^2 + (latA - latD)^2 < (longB - longD)^2 + (latB - latD)^2

longB = longA;
latB = latA;
end

longG = abs(longB - longC)/2 + min(longB,longC);

latG = abs(latB - latC)/2 + min(latB,latC);
longPath = [longPath, longG];
latPath = [latPath, latG];
plot(longIn, latIn, longOut, latOut, longABC, latABC, longPC, latPC, longPath,
latPath);

vertN = 5;
vertM = size(longVert) + 1;
evalResponse = input('prompt');
while vertN < vertM(1,2)
longA = longC;
latA = latC;
longC = longD;
latC = latD;
longD = longVert(1,vertN);
latD = latVert(1,vertN);
vertN = vertN + 1;
longP = longG;
latP = latG;

longABC = [longA, longB, longC , longA];

latABC = [latA , latB , latC , latA ];
longPC = [longP, longC];
latPC = [latP , latC ];
longPD = [longP, longD];
latPD = [latP , latD ];

if (longA - longD)^2 + (latA - latD)^2 < (longB - longD)^2 + (latB - latD)^2

longB = longA;
latB = latA;
end

longG = abs(longB - longC)/2 + min(longB,longC);

latG = abs(latB - latC)/2 + min(latB,latC);
longPath = [longPath, longG];
latPath = [latPath, latG];

plot(longIn, latIn, longOut, latOut, longABC, latABC, longPC, latPC,

longPath, latPath);
evalResponse = input('prompt');
end

longP = longG;
latP = latG;
longA = longC;
latA = latC;
longC = longD;
latC = latD;
9
longABC = [longA, longB, longC , longA];
latABC = [latA , latB , latC , latA ];
longPC = [longP, longC];
latPC = [latP , latC ];
longPD = [longP, longD];
latPD = [latP , latD ];

longG = abs(longA - longC)/2 + min(longA,longC);

latG = abs(latA - latC)/2 + min(latA,latC);
longPath = [longPath, longG];
latPath = [latPath, latG];

plot(longIn, latIn, longOut, latOut, longABC, latABC, longPC, latPC, longPath,

latPath);
evalResponse = input('prompt');
plot(longIn, latIn, longOut, latOut, longPath, latPath);

Appendix D

Figure 1: Depiction of Object Avoidance Algorithm

Appendix E

Figure 1: Connection Diagram of rover components

10